The clinical actionability and evolution of mutational processes in metastatic cancer (2020)
Cancers are characterized by somatic mutation arising from the interplay of mutagen exposure and deficient DNA repair. Whole genome sequencing of tumours reveals characteristic patterns of mutation, known as mutation signatures, which often correspond with specific processes such as cigarette smoke exposure or the loss of a DNA repair pathway. Quantifying DNA repair deficiency can have clinical implications. Cancer chemotherapies which induce DNA damage are known to be more effective against cancers with deficient DNA repair. However, it is not yet known whether mutation signatures can serve as reliable predictive biomarkers for response to these treatments. Furthermore, the current understanding of mutation signatures stems largely from studies of primary, untreated tumours, whereas metastasis underpins as much as 90% of cancer-related mortality. This thesis aims to (1) describe the association between mutation signatures and clinical response to DNA damaging chemotherapy, (2) enable accurate personalized assessment of mutation signatures and their evolution over time, and (3) characterize the evolution of mutational processes in metastatic cancers. To assess clinical actionability, we quantified signatures of single nucleotide variants, structural variants, copy number variants, and small deletions in 93 metastatic breast cancers, 33 of which received platinum-based chemotherapy. We found that patients with signatures of homologous recombination deficiency had improved responses and prolonged treatment durations on platinum-based chemotherapy. Next, we formulated a Bayesian model called SignIT, which improves the accuracy of individualized mutation signature analysis and infers signature evolution over tumour subpopulations. We demonstrated SignIT’s superior accuracy on both simulated data and somatic mutations from The Cancer Genome Atlas, and validated temporal dissection using whole genomes from 24 multiply-sequenced cancers. We highlighted a potential clinical application of mutation in a BRCA1-mutated pancreatic adenocarcinoma with low Homologous Recombination Deficiency (HRD) signature but exceptional response to platinumcontaining chemotherapy. Finally, we deciphered mutation signatures from nearly 500 metastatic cancer whole genomes, revealing evolution of mutational processes associated with late metastasis and exposure to cytotoxic chemotherapy. Taken together, our findings demonstrate the complex interplay of factors shaping the metastatic cancer genome. We highlight both clinical opportunities of studying genomic instability and the additional insights available from understanding their temporal evolution.